Papers of the day   All papers

Green AI

Comments

hardmaru: Green AI 🌳 “We want to shift the balance towards the Green AI option — to ensure that any inspired undergraduate with a laptop has the opportunity to write high-quality papers that could be accepted at premier research conferences.” 👩🏻‍💻 https://arxiv.org/abs/1907.10597 https://t.co/L8e1EqiQXD

8 replies, 586 likes


Andrej Karpathy: 💻🧠+🌍🌳 recent reads: Green AI vs Red AI https://arxiv.org/abs/1907.10597 and "Tackling Climate Change with Machine Learning" https://www.reddit.com/r/MachineLearning/comments/da30mv/r_tackling_climate_change_with_machine_learning/ https://t.co/8yRj6g7HIr

9 replies, 548 likes


Roy Schwartz: The focus on SOTA has caused a dramatic increase in the cost of AI, leading to environmental tolls and inclusiveness issues. We advocate research on efficiency in addition to accuracy (#greenai). Work w/ @JesseDodge @nlpnoah and @etzioni at @allen_ai https://arxiv.org/abs/1907.10597

1 replies, 159 likes


Taco Cohen: Green AI: "[Deep Learning] computations have a surprisingly large carbon footprint. [...] This position paper advocates a practical solution by making efficiency an evaluation criterion for research along-side accuracy and related measures" https://arxiv.org/abs/1907.10597

3 replies, 77 likes


samim: The greenest form of "AI" is "No AI". Let's ask ourselves earnestly, does this task really need to be computerized & automated - or is it possibly healthier & saner for humans, flora & fauna to do the task? There is no free lunch - every augmentation leads to amputation.

3 replies, 76 likes


Leo Dirac: I don't see a fundamental problem using massive compute to advance AI - research means pushing what's possible with today's tech to inform tomorrow. But I fully support the proposal by @etzioni and others to publish compute cost and efficiency in papers.

4 replies, 31 likes


Tim Miller: Great initiative: use energy efficiency as an evaluation metric for ML research. Focus more on new ideas than larger computing resources.

0 replies, 23 likes


Tom Simonite: "We propose reporting the financial cost or 'price tag' of developing, training, and running [machine learning] models" — @etzioni @JesseDodge @nlpnoah @royschwartz02 say researchers should disclose how much computing power they use to encourage greener AI https://arxiv.org/abs/1907.10597

0 replies, 15 likes


Parisa Rashidi: The amount of compute used to train deep learning models has increased 300,000x in 6 years. The community should take into account not just AUC and precision, but also the carbon footprint. Paper by @etzioni @nlpnoah et al. https://arxiv.org/abs/1907.10597 https://t.co/HXUVV8aOQx

1 replies, 14 likes


Ethem Alpaydın: Green AI: "...making efficiency an evaluation criterion for research alongside accuracy and related measures." https://arxiv.org/abs/1907.10597

0 replies, 14 likes


Ulli Waltinger: #GreenAI needs to be an emerging trend as the computations required for #deeplearning research have been doubling every few months, resulting in an estimated 300,000x increase from 2012 to 2018 #AISustainabilityQuest https://arxiv.org/abs/1907.10597 https://t.co/usSD3bPfgV

0 replies, 11 likes


Ant Kennedy: Interesting read from @allen_ai https://arxiv.org/abs/1907.10597 around introducing efficiency as an evaluation criterion for your models. Something we should all be considering especially given training a model can produce more CO₂ than an a lifetime of car travel 😨 #GreenAI

1 replies, 11 likes


Ant Kennedy: @DanLarremore @anne_e_currie There are some interesting ideas emerging around using efficiency as a metric when training https://arxiv.org/abs/1907.10597

0 replies, 11 likes


Paul Strobel: The director of AI at @Tesla tweeted about the summary i made about the paper "Tackling Climate Change with Machine Learning". Thank you @karpathy for giving more awareness to the topic. Happy to get in touch & share thoughts 🌳🌍💡

0 replies, 10 likes


Peter Steinbach: For me, a long awaited and much needed initiative. Can't wait to read the paper. The abstract already made my eyes wet with anxiety. #Green500

0 replies, 8 likes


Geraint Rees: Green AI - interesting new preprint https://arxiv.org/abs/1907.10597 via the brilliant @ExponentialView

0 replies, 5 likes


Ada Lovelace Institute: 3⃣ There are tools/tips for data scientists to start reducing climate impact: 🔹Datacentre choice @coedethics: https://docs.google.com/document/d/1eCCb3rgqtQxcRwLdTr0P_hCK_drIZrm1Dpb4dlPeG6M/edit 🔹 Estimate ML model emissions #mlco2: https://mlco2.github.io/impact/ 🔹 Energy efficiency as research evaluator: https://arxiv.org/abs/1907.10597 #GreenAI

1 replies, 5 likes


Séb 🌐: The amount of compute used to train deep learning models has increased 300,000x in 6 years, leaving a large carbon footprint. This paper advocates making 'efficiency' an evaluation criterion for research alongside accuracy and related measures: https://arxiv.org/pdf/1907.10597.pdf https://t.co/NWNbEMgACh

1 replies, 5 likes


Luca Soldaini: Very nice proposition paper; using # of floating point operations seems to be a decent standard to approximate computational cost of a model, although I'm not convinced it captures the intuitive notion of "efficiency", which is imho "do the best with the resources you have".

1 replies, 4 likes


Titus von der Malsburg: Paper advocating for green AI that takes energy efficiency into account: https://arxiv.org/abs/1907.10597

0 replies, 4 likes


Victor Sanh: If I had to pinpoint ONE reading for this we: Green AI (https://arxiv.org/pdf/1907.10597.pdf) by @royschwartz02 @JesseDodge @nlpnoah @etzioni. It advocates that efficiency should be a metric we chase just like accuracy.

1 replies, 4 likes


Jim Schwoebel: https://www.technologyreview.com/s/613630/training-a-single-ai-model-can... something to be conscious of when working on models! https://arxiv.org/pdf/1907.10597.pdf -- Alle.. (sent from @Protea_app)

0 replies, 4 likes


Meg White: In today's batch of ML papers: "The computations required for deep learning research have been doubling every few months, resulting in an estimated 300,000x increase from 2012 to 2018. [...] Our goal is to make AI both greener and more inclusive" https://arxiv.org/abs/1907.10597

0 replies, 4 likes


Giorgio Patrini 🛡️👾: Let's hope this becomes a by-default reference in empirical papers in ML It's easy to misunderstand why this is important. *It isn't* about limiting comp resources for empirical research. It's about giving guidelines to practitioners who'll train those models thousands of time

0 replies, 4 likes


Victoire Louis: And here is the paper @MrsCaroline_C thanks for the recommandation https://arxiv.org/pdf/1907.10597.pdf Machine Learning is great ... and what about the environment ? #ClimateChange #MachineLearning #Innovation

0 replies, 4 likes


Josie Young [🏳️‍🌈 ally]: @johnchavens @deeplearningai_ newsletter recently had a report from @allen_ai on measuring the carbon footprint of AI - really useful and a great start for building a green AI toolset! https://arxiv.org/abs/1907.10597

0 replies, 4 likes


anjali 🚀 (but scary 🎃): New research from @allen_ai gives the ML community concrete ways to change how we measure success -- benchmarking energy consumption will also help us develop more accessible models See "Green AI" by @royschwartz02 @JesseDodge @nlpnoah @etzioni https://arxiv.org/pdf/1907.10597.pdf

1 replies, 3 likes


Anna Rogers: The #AcademicTwitter #GreenAI panel from other threads and papers: * @nlpnoah @royschwartz02 @JesseDodge @etzioni https://arxiv.org/abs/1907.10597 * @strubell @andrewmccallum https://arxiv.org/abs/1906.02243 * @alex_lacoste_ @vict0rsch https://arxiv.org/abs/1910.09700 * @bkbrd

1 replies, 3 likes


Beliz Gunel: Focusing *only* on SOTA accuracy gives a clear edge to organizations that have the resources to train these fat models, creating barriers to participation in certain types of NLP research. Great review!

0 replies, 3 likes


Poilvet Eric: The computations required for deep learning research have been doubling every few months, resulting in an estimated 300,000x increase from 2012 to 2018. #Green #AI is a must. Interesting article via ⁦@deeplearningai_⁩ https://arxiv.org/abs/1907.10597?utm_campaign=The%20Batch%20081419%20MLY%20Intro&utm_source=hs_email&utm_medium=email&utm_content=75698304&_hsenc=p2ANqtz-8fEnmueToUyiSHgzO_7c6p_XlDqKxUhPsyNsbE-Z-C7K-NkUjeqCCKMBu2-ln9mo4RQ7-KPzWCeVC6gi41ia5cQamosA&_hsmi=75698304

0 replies, 2 likes


digital dynamics: - Green AI - "This position paper advocates a practical solution by making efficiency an evaluation criterion for research alongside accuracy and related measures." https://arxiv.org/abs/1907.10597 #ethicalAI #responsibleAI #AI

0 replies, 2 likes


Spencer Dixon 🤖: I’ve talked a lot about tech giants and how in the future they’ll move into the #conservation world. Here’s yet another step in that direction. Google, Microsoft and Deepmind’s recent paper ‘Tackling Climate Change with Machine Learning’ https://arxiv.org/abs/1907.10597

0 replies, 2 likes


Noah Smith: #greenai

0 replies, 2 likes


Debanjan Mahata: A report (https://arxiv.org/pdf/1907.10597.pdf) from the @allen_ai argues that, for new models, energy efficiency is as important as accuracy. The report sets forth several ways to assess AI's carbon emissions. #AI #MachineLearning #NLProc

0 replies, 1 likes


BISHAL SANTRA: Good to see someone's starting to bother about this side of spectrum also. Green AI The computations required for deep learning research have been doubling every few months, resulting in an estimated 300,000x increase from 2012 to 2018.https://arxiv.org/abs/1907.10597 #NLProc @arXiv_Daily

0 replies, 1 likes


Romain Rivière: Computations required for deep learning research have been doubling every few months, resulting in an estimated 300,000x increase from 2012 to 2018... https://arxiv.org/pdf/1907.10597.pdf #GreenAI #NLP #efficiency

0 replies, 1 likes


Kostas Stathoulopoulos: @CassieRobinson That's why the energy efficiency of developing, training and running a model should be an evaluation criterion of research: https://arxiv.org/abs/1907.10597

0 replies, 1 likes


David Schatsky: The computations required for deep learning research have been doubling every few months, resulting in an estimated 300,000x increase from 2012 to 2018 https://arxiv.org/pdf/1907.10597.pdf

0 replies, 1 likes


Alex J. Champandard: 8/ Metrics. Let's start measuring, reporting and comparing training times in papers. Beyond that, is there any interest in a semi-automated competition to create "Green" models whose innovation does not merely rely on more data or compute? 🌱 https://arxiv.org/abs/1907.10597

0 replies, 1 likes


Felipe Ducau: Monster-size models are *one* possible way to go about getting better results. Green AI proposes to pay attention not only to the optimization metric but the cost of getting at it. Smaller, faster, more data-efficient models are just better. http://arxiv.org/abs/1907.10597

0 replies, 1 likes


aDynamics: #Green #AI https://arxiv.org/abs/1907.10597 #arXiv

0 replies, 1 likes


JuliaGo: Curious to see what the findings of our participants will be after the next 3 days @ the #nexushackathon where we will have one track regarding #co2 footprint of deep learning 👀 @HackathonNexus

0 replies, 1 likes


Brian Merchant: "researchers should disclose how much computing power they use to encourage greener AI" Yep -- this is going to become a major issue as computationally heavy AI continues to ramp up and big tech builds out server farms to keep pace

0 replies, 1 likes


Ada Lovelace Institute: 3⃣ There are tools/tips for data scientists to start reducing climate impact: 🔹Datacentre choice @coedethics: https://docs.google.com/document/d/1eCCb3rgqtQxcRwLdTr0P_hCK_drIZrm1Dpb4dlPeG6M/edit 🔹 Estimate ML model emissions #mlco2: https://mlco2.github.io/impact/ 🔹 Energy efficiency as research evaluator: https://arxiv.org/abs/1907.10597 #GreenAI

1 replies, 0 likes


Content

Found on Jul 26 2019 at https://arxiv.org/pdf/1907.10597.pdf

PDF content of a computer science paper: Green AI